Design Of A Cattle Body Measurement System Based On Kinect V4

Growth parameter of cattle body size is an important indicator that measures the growth of grassland animal husbandry cattle,which can be used to visually reflect the growth and development of cattle.The traditional method of cattle body size measurement is artificial contact measurement,which is expensive,and accompanied by a risk that cattle attack peo-ple.In response to this problem,combined with computer vision,depth learning,image processing,etc.,and based on Kinect V4 sensor,a system of measuring cattle body size was designed and implemented.First,color and depth image information of the side of cattle body was collected by Kinect V4 sensor,and the information was fused into RGB-D image.Second,a CABM2022 dataset of feature parts was created,and deep learning convolutional neural network YOLOv5(You Only Look Once v5)target detection algorithm was used to detect the color image in-formation parts of the RGB-D image,the position information of feature parts of the body size in the RGB-D image was obtained,in order to reduce interference of other parts of cat-tle body and background on the detection of cattle body size measurement points.Third,according to the position information of feature parts,acquired image can be determined whether it is suitable for the body size measurement,and the suitable image information was cropped.Combined with Canny edge detection,contour extraction,and other image pro-cessing algorithms in Open CV(Open source Computer Vision)image processing library,and a feature that depth image is not easy to be disturbed by the outside world,the edge pro-file of the cattle measuring point was extracted in the cropped RGB-D partial image.And then,cattle measuring points were extracted in the edge profile by the polynomial fitting and three-point arc curvature method,and the points were converted back to the correspond-ing position in the RGB-D image.Then,the depth information of the RBG-D image was converted to two-dimensional image,and the ground information was converted to three-dimensional coordinate system.Combined with random sampling algorithm and ground in-formation,a ground plane expression was fitted in three-dimensional coordinate system,and a calculation method of the body size was designed under three-dimensional coordinate sys-tem by three-dimensional European distance and point-to-line distance formula.Finally,the graphical user interface is designed by Qt Designer,and the body size measurement system is designed based on the PYQT framework.After the experimental measurement when the sensor and the side of the model yellow cattle form differential deflection angles in the laboratory and the field measurement of the pasture cattle in Kunbei pasture of Baotou,the average relative error of the cattle size ruler data is less than 2% and 5% respectively.The experimental results show that the measure-ment accuracy of the body size measurement method is high,which can meet the demand for the growth parameter measurement of cattle body size in natural state.This dissertation pro-vides a new method for measuring growth parameters of cattle body size based on computer visual.